Power Systems: Look Ahead, Don’t Look Back
April 21, 2014 Timothy Prickett Morgan
There is no two ways about it. The declining sales of IBM‘s Power Systems line is a cause for concern at the company and is as troubling in some ways as the drop in the company’s sales of X86-based machinery and various kinds of storage arrays. Many IBM i shops buy all of this gear from Big Blue, and they may be wondering what kind of commitment IBM has for the long term with any of these platforms, given the diminishing revenues.
The fact of the matter is that IBM not only needs to stay in the hardware business to be relevant, but that the company is committed to doing so–with the caveat that it is no longer interested in pushing commodity tin when it can peddle sophisticated big iron in the form of sophisticated Power Systems clusters or big NUMA machines based on Power or System z processors. To a certain extent, once IBM left the PC business a decade ago, it was only a matter of time before it would be under pressure in the X86 server business. Those vendors who buy the most from Intel, including a mix of PC and server chips, get the best prices and therefore can either sell for less or get more margin or a little of both. Hewlett-Packard and Dell both have significantly higher X86 server shipments than does IBM, and they have PC businesses to boot. So there was no way IBM was going to be able to compete. That is why IBM is selling off the System x server business to Lenovo for $2.3 billion. And even that might not be enough if Power processors take off in hyperscale datacenters and supercomputing centers, particularly if Intel lowers the pricing boom as, I might point out, an IBM circa 1970 might have done to anyone who threatened its mainframe business.
The thing to remember is that even as IBM has trimmed down the Power Systems employee pool, the company is focusing on different parts of the market where Power chips do have an advantage over X86 and ARM processors, and as I have said many times before, anything that lets Power stronger makes IBM i live longer. This, ultimately for users of IBM i, is what matters.
“The repositioning of the Power platform, the announcement of Power8, new announcements in storage, and the right-sizing of the business will all contribute, as we move through the rest of the year,” explained Martin Schroeter, IBM’s chief financial officer, in a conference call with Wall Street analysts going over the financial results for the first quarter of 2014. “This, together with the divestiture of System x, will result in a smaller and more profitable hardware business going forward. IBM will remain a leader in high-performance and high-end systems, in storage and in cognitive computing, and we will continue to invest in R&D for advanced semiconductor technology.”
I know IBM is keen on turning Watson into a cloudy services business running on Power-based machines in its SoftLayer clouds, and because I want the Power chip to survive in the datacenter after it has been vanquished from desktops, game consoles, and plenty of other embedded applications, I stand behind that effort. But I don’t think Watson services are going to drive Power chip volumes. It takes a lot less iron to run Watson today that it did for the Jeopardy! game show, more than can be accounted for by Moore’s Law improvements from Power7 to Power8. IBM has learned how to tune the Watson question-answer software as well as expand its capability. What took racks and racks of machines to run can now be done by three rack servers–soon to be two with Power8.
This is why IBM is so keen on the OpenPower Foundation and licensing the technology in its Power Systems to an ecosystem of partners who build Power-based machines. And some day, it could turn out that you actually run IBM i on a third-party machine, much as you run Windows or Linux from Microsoft or Red Hat on an X86 box from a number of different vendors. IBM can make the chip, or design the chip and let someone else make it, and then make the money on AIX and IBM i. And frankly, I think this is precisely what IBM truly desires, even though it has not said as much publicly. But it is no coincidence perhaps that a motherboard maker (Tyan), a chip maker (Suzhou PowerCore), and a homemade server maker (Google) were among the first members of the OpenPower Foundation.
I know many of us were ready to build our own Power-based machines many times over the past 15 years because we thought we could do a better job. We may yet get our chance.
In the first quarter ended in March, Power Systems machines logged their seventh quarter of declines out of the past eight, and the eighth of those (in the third quarter of 2012) was flat so that was not much to write home about, either. In Q1, Power Systems revenues were down 22 percent, and IBM did not elaborate about what was selling and what was not, but IBM has been pretty clear that it thought it had too many people for the diminished revenue stream of Power-based gear. Schroeter also cautioned Wall Street–and by extension, Power Systems customers and business partners–from thinking there was a quick fix to getting the Power business stabilized and growing:
“Power is not going to turn around right away, and in fact I think 2014 is going to be a difficult year for Power if you look at year-to-year revenue growth. But we are absolutely taking the right actions to make that Power platform a sustainable and a highly attractive economic model for us. We have taken cost out and we are getting the right technology and ecosystem around it to make sure it has a long-term future.”
The System z mainframe business had a 40 percent revenue drop in the March quarter, and the aggregate MIPS shipped to customers was down by 19 percent compared to the year-ago period, Schroeter said. But, he pointed out, if you compare the System z11 line to the current System z12 line, the aggregate amount of MIPS in the base is 26 percent higher with the z12 than it was at the same point in the z11 cycle, and this is what IBM is really looking at every day and week and month. That is because mainframe software is rented monthly. And that means IBM is collecting more mainframe software revenue, or has room to cut prices by that increase and keep its revenues stable. (I suspect the latter is happening, given that its “other middleware” segment in Software Group had flat sales in the quarter.) System x sales were down 18 percent in the quarter, which is no wonder at all with the System x sale in process and customers uncertain about the future despite all of the assurances both IBM and Lenovo have given. Moreover, both Hewlett-Packard and Dell are slogging it out against each other in the hyperscale datacenters, with the likes of Quanta and Supermicro also putting immense downward pressure on X86 server prices. IBM’s storage business is not doing well, particularly at the high-end, with sales across all disk, flash, and tape products down 23 percent. IBM probably does a lot of storage business as part of system upgrades, so again this is not a big surprise.
Systems and Technology Group had sales of $2.39 billion, down 23 percent, and posted a pre-tax loss of $660 million. About $200 million of that was driven by charges related to layoffs that IBM made during the first quarter. IBM had a $405 million pre-tax loss in the year-ago period against $3.11 billion in revenues, so it is a wonder that the losses were not higher. The fact that mainframe margins are significantly higher at this point in the 12 quarter cycle for the machines helps; all IBM is doing for most customers is activating cores, and that costs nothing. Building the mainframes is another story entirely. Server revenues came to $1.58 billion, down 24 percent, and storage sales were $502 million, down 23 percent. The IBM hardware systems business is now about half what it was 25 years ago, just to give you some perspective.
Software Group had $5.66 billion in sales for Q1, up 1.6 percent, and pre-tax income fell 4.7 percent to $1.92 billion. WebSphere has become a very broad category now and sales are up 12 percent. Database and related tools sales were up 1 percent, Tivoli security and systems management software rose 7 percent, and Rational development tools rose 1 percent. Workforce Solutions, the part of IBM that used to be the Notes/Domino stack and now includes other wares, was off 4 percent. These key branded middleware products, as IBM calls them, accounted for $3.68 billion in revenues and rose by 4 percent in aggregate. The other middleware accounted for a little over $1 billion in sales and was flat; operating systems brought in $510 million and were down nearly 9 percent. Software Group had a $200 million write off related to the workload rebalancing during the quarter.
Global Services had a 2 percent shrink in the quarter, to $14.1 billion and yes, is nearly ten times as large as IBM’s systems hardware business. Pre-tax income here was down nearly 14 percent to $1.97 billion, and around $600 million in writeoffs from the layoffs pushed that down.
Across the whole company, Big Blue’s revenues fell by 3.9 percent, and net income was down 21.4 percent to $2.38 billion. Without the writeoffs driven by layoffs, IBM’s net income would have been up nicely, but it would not be positioning for the future, where it needs to extract many billions of dollars for dividends and share buybacks to meet its 2015 targets.